PhD Dissertation, EPFL
This research exemplifies a human-centric approach to visual interest prediction, using architectural case studies to evaluate daylight composition from a fixed view position across an immersive range of view directions. This full 360° span replicates the daylight distribution perceived at eye-level across the entire field-of-view, using High Dynamic Range (HDR) renderings with a wide fisheye specification as input for the modified spatial contrast (mSC) metric developed by the authors and published in Lighting Research & Technology. From this fixed view position, the time of day and day of year are varied so as to evaluate dynamic conditions for two selected case studies across an occupant’s immersive point-of-view. This occupant-driven approach does not seek to replace existing methods for evaluating illumination, but rather propose a human-centric approach for analyzing those aspects of daylight that are dependent on light received at the eye level. This model will help shift our attention from spatially-dependent (as in task plane illumination or fixed field-of-view comfort) to human-centric performance assessment methods which account for an occupants’ visual immersion in daylit architectural space.
When we transition from a fixed view position such as an 80° x 60° rectangular rendering to a human-centric approach for predicting visual interest, three main challenges emerge: 1) how can we use image based metrics like Spatial Contrast (developed for 2D images) to predict visual interest across a 3-dimensional space and 2) how can we quantify the variation in this prediction based on dynamic environmental conditions and view direction (i.e. across the entire 360° view range). Finally, how can these predictions across a subject’s field-of-view, over time, and across sky conditions help us to create a holistic assessment of the indoor environment as it impacts programmatic use and design intent. In short, what does this prediction tell us about the occupant’s experience in the architecture we are assessing? This work has been submitted to Building & Environment.